a single
shown.
)lained,
d. The
'eneous
models
leity as
nt GIS
g their
r-based
proach.
tics of
IS. For
objects
to the
mental
tabase.
zeneity
model
curs in
les are
y a set
onflict
ne real
s. Data
lefined
object
cts are
uctural
( of an
it as a
tabase
They
eity to
three
e and
sneity,
flict.
fferent
lel Vs
elated
, eg.
red as
of the
; and
ferent
serve
nantic
S for
[S for
topographic mapping.
An architecture for sharing data stored in disparate geographic
information systems is shown in Figure 4. It is a four levels
architecture:
e Atthe first level, level 1, schemes are representing the part of
the database that each member in the federation is willing to
share. These schemes are represented with the native
language of the host.
e The second higher level of schemes is to model the same data
posted at the first level but with a uniform data model (e.g.
Object Oriented).
e The third level is a global schema which provides a uniform
perspective of the information posted at the second level.
e The federated schemes are supporting different scale levels,
i.e., national, regional, and local.
e Another fourth level of schemes is representing the
application specific views which can be partly retrieved form
the federation (e.g. application database).
The object oriented sharable schemes in level 2, Figure 4, are
represented in the local servers, Figure 2. The federated object
oriented schemes, Figure 4, are represented in their
corresponding global servers GS;, Figure 2. The multi-level
global server in Figure 4, is the one which provides the link
between the applications, i.e., level 4 in Figure 2.
As shown in Figure 5, there could be several watershed
management projects running in parallel. Each individual
project can have positive impact when viewed locally. However,
they might have negative impact on the basin, when their overall
impact is screened. For this reason management projects
proposed at watershed level should be assessed at the basin
level. The implementation of the system architecture proposed
in Section 5 aimed at showing this case. Management practices
where introduced into watersheds. The management practices
proved to minimize the soil erosion. This impact was quantified
using AGNPS model. The impact of these management practices
was then analyzed at the basin level. This impact was quantified
using DUFLOW. In order to achieve this objective, the multi-
level decision support system should allow data and decision
transfer.
In the previous sections the system architecture of the multi-
level decision support system for watershed management is
shown. Moreover, an architecture for resolving the aspects of
heterogeneity in the databases supporting such system was
proposed in this section. A prototype which implements the
system is presented in the next section. Only the system
components and functionality are shown. The supporting data
models are outside the scope of this paper.
5. SYSTEM IMPLEMENTATION
Four main software packages were used to implement MLSDSS
the client-server operations:
* Nexpert Object as the object-oriented shell and rule-base
system.
* Arc/Info as the GIS platform.
669
e. AGNPS as the erosion and water quality simulation model at
the regional level, i.e., catchments and subcatchments.
e Duflow as the erosion and water quality simulation model at
the national level, i.e., basin. It is used to analyse the overall
impact of management practices on the basin.
Nexpert Object supports a rich range of representation features.
In Nexpert, the domain is modelled in terms of objects, classes,
and properties. Rules are used to manipulate the objects and
class structures. The specific properties of objects and classes
are called slots. Meta-slot attributes are used to describe certain
characteristics of the slot. Nexpert Object shell is used to build
an object oriented shell around the sharable data and represent
the object network for the client, local servers, and global
Tourism A
t 7 Water Management
Soil Conservation
Agriculture
Overall Impact
Figure 5 Several management projects in one basin
server. The rule-base is used to design the global server and the
database bridge will be used by the local servers to transfer data
between dBase and Nexpert representation. Nexpert object will
have several roles in our system. First it is used for integration
and implement the components of the SDSS, i.e., ARC/Info,
AGNPS, and Duflow, [Espinoza E., 1995 and; Mabote T.,
1995] gave a detailed description of this system. Second
Nexpert is used for developing the schemes of the MLDSS and
as a mediator for resolving the heterogeneity between the
supporting databases. In this paper we only emphasis the
technique followed for implementing the servers.
Servers Implementation
To simulate the servers in the proposed architecture,
corresponding Nexpert's Knowledge Bases are developed, one
for each server.
LOCAL SERVER
Using NEXPERT, an object oriented shell is put on top of the
local database. This shell is basically the object oriented view of
the sharable data schema using the canonical or common data
model. To provide the server functionality it is necessary to
develop methods that enable the communication between the GS
and the Local Database. Using these methods, messages can be
passed from the GS to the LS in order to retrieve data, as
requested by the client, from the underlying database. The
retrieval asked by the GS is originally an Object Oriented query
that the LS has to translate into an equivalent query for the local
DBMS.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996